US20080051930A1 - Scheduling method for processing equipment - Google Patents

Abstract

Methods and apparatus for improving substrate throughput prior to processing a plurality of substrates in a cluster tool having a plurality of processing chambers and at least two robots are provided. The method comprises developing a process sequence, storing the process sequence on a storage device, and transferring the process sequence from the storage device to a controller in communication with the cluster tool. The process sequence comprises depositing one or more uniform resist layers on the surface of the plurality of substrates, transferring the plurality of substrates out of the cluster tool to a separate stepper or scanner tool to pattern the surface of the plurality of substrates by exposing the resist layer to a resist modifying electromagnetic radiation, and then developing the patterned resist layer.

Description

CROSS-REFERENCE TO RELATED APPLICATIONS
• This application claims benefit of U.S. Provisional Patent Application No. 60/806,906, filed Jul. 10, 2006, which is herein incorporated by reference.
BACKGROUND OF THE INVENTION
• 1. Field of the Invention
• Embodiments of the invention as recited in the claims generally relate to an integrated processing system containing multiple processing stations and robots that are capable of processing multiple substrates in parallel. In particular, the invention relates to scheduling methods for an integrated processing system.
• 2. Description of the Related Art
• The process of forming electronic devices is commonly done in a multi-chamber processing system (e.g., a cluster tool) that has the capability to sequentially process substrates, (e.g., semiconductor wafers) in a controlled processing environment. A typical cluster tool used to deposit (i.e., coat) and develop a photoresist material, commonly known as a track lithography tool, will include a mainframe that houses at least one substrate transfer robot which transports substrates between a pod/cassette mounting device and multiple processing chambers that are connected to the mainframe. Cluster tools are often used so that substrates can be processed in a repeatable way in a controlled processing environment. A controlled processing environment has many benefits which include minimizing contamination of the substrate surfaces during transfer and during completion of the various substrate processing steps. Processing in a controlled environment thus reduces the number of generated defects and improves device yield.
• The effectiveness of a substrate fabrication process is often measured by two related and important factors, which are device yield and the cost of ownership (CoO). These factors are important since they directly affect the cost to produce an electronic device and thus a device manufacturer's competitiveness in the market place. The CoO, while affected by a number of factors, is greatly affected by the system and chamber throughput, or simply the number of substrates per hour processed using a desired processing sequence. A process sequence is generally defined as the sequence of device fabrication steps, or process recipe steps, completed in one or more processing chambers in the cluster tool. A process sequence may generally contain various substrate (or wafer) electronic device fabrication processing steps. In an effort to reduce CoO, electronic device manufacturers often spend a large amount of time trying to enhance the process sequence and chamber processing time to achieve the greatest substrate throughput possible given the cluster tool architecture limitations and the chamber processing times. In track lithography type cluster tools, since the chamber processing times tend to be rather short, (e.g., about a minute to complete the process) and the number of processing steps required to complete a typical process sequence is large, a significant portion of the time it takes to complete the processing sequence is taken up transferring the substrates between the various processing chambers.
• A typical track lithography process sequence will generally include the following steps: depositing one or more uniform photoresist (or resist) layers on the surface of a substrate, then transferring the substrate out of the cluster tool to a separate stepper or scanner tool to pattern the substrate surface by exposing the photoresist layer to a photoresist modifying electromagnetic radiation, and then developing the patterned photoresist layer. If the substrate throughput in a cluster tool is not robot limited, the longest process recipe step will generally limit the throughput of the processing sequence. This is usually not the case in track lithography process sequences, due to the short processing times and large number of processing steps. Typical system throughput for the conventional fabrication processes, such as a track lithography tool running a typical process, will generally be between 100-120 substrates per hour.
• Other important factors in the CoO calculation are the system reliability and system uptime. These factors are very important to a cluster tool's profitability and/or usefulness, since the longer the system is unable to process substrates the more money is lost by the user due to the lost opportunity to process substrates in the cluster tool. Therefore, cluster tool users and manufacturers spend a large amount of time trying to develop reliable processes, reliable hardware, and reliable systems that have increased uptime.
• The push in the industry to shrink the size of semiconductor devices to improve device processing speed and reduce the generation of heat by the device, has caused the industry's tolerance to process variability to diminish. Due to the shrinking size of semiconductor devices and the ever increasing device performance requirements, the allowable variability of the device fabrication process uniformity and repeatability has greatly decreased. To minimize process variability an important factor in the track lithography processing sequences is the issue of assuring that every substrate run through a cluster tool has the same “wafer history.” A substrate's wafer history is generally monitored and controlled by process engineers to assure that all of the device fabrication processing variables that may later affect a device's performance are controlled, so that all substrates in the same batch are always processed the same way. To assure that each substrate has the same “wafer history” requires that each substrate experiences the same repeatable substrate processing steps (e.g., consistent coating process, consistent hard bake process, consistent chill process, etc.) and the timing between the various processing steps is the same for each substrate. Lithography type device fabrication processes can be especially sensitive to variations in process recipe variables and the timing between the recipe steps, which directly affects process variability and ultimately device performance.
• Therefore, there is a need for a system, a method, and an apparatus that can process a substrate so that it can meet the required device performance goals and increase the system throughput while maintaining a constant “wafer history” and thus reduce the process sequence CoO.
SUMMARY OF THE INVENTION
• Embodiments of the invention as recited in the claims generally provide a method for processing substrates using a multi-chamber processing system (e.g., a cluster tool) that has an increased system throughput and repeatable wafer processing history.
• In one embodiment a method of improving substrate throughput prior to processing a plurality of substrates in a cluster tool having a plurality of processing chambers and at least two robots is provided. The method comprises planning a schedule and the corresponding actions performed by each component of the cluster tool, storing the schedule on a storage device, and transferring the schedule from the storage device to a controller in communication with the cluster tool.
• In another embodiment, a method for improving substrate throughput prior to processing a plurality of substrates in a cluster tool having a plurality of processing chambers and at least two robots is provided. The method comprises developing a process sequence, storing the process sequence on a storage device, and transferring the process sequence from the storage device to a controller in communication with the cluster tool. The process sequence comprises depositing one or more uniform resist layers on the surface of the plurality of substrates, transferring the plurality of substrates out of the cluster tool to a separate stepper or scanner tool to pattern the surface of the plurality of substrates by exposing the resist layer to a resist modifying electromagnetic radiation, and then developing the patterned resist layer.
• In yet another embodiment, a system for processing a substrate is provided. The system comprises a cluster tool having a plurality of processing chambers and at least two robots, a controller in communication with the cluster tool, a memory coupled to the controller, an offline server, a storage device coupled to the offline server, and a memory coupled to the offline server. The memory comprises a computer-readable medium having a computer readable program embodied therein for improving the throughput of substrates in the cluster tool. The computer readable program comprises computer instructions for developing a recipe for improving substrate throughput in the cluster tool by controlling the movements of the at least two robots. The computer instructions comprise planning a schedule and the corresponding actions performed by each component of the cluster tool, storing the schedule on the storage device, and transferring the schedule from the storage device to the controller.
BRIEF DESCRIPTION OF THE DRAWINGS
• So that the manner in which the above recited features of the present invention can be understood in detail, a more particular description of the invention, briefly summarized above, may be had by reference to embodiments, some of which are illustrated in the appended drawings. It is to be noted, however, that the appended drawings illustrate only typical embodiments of this invention and are therefore not to be considered limiting of its scope, for the invention may admit to other equally effective embodiments.
• FIG. 1A is an isometric view illustrating a processing system in accordance with one embodiment of the present invention;
• FIG. 1B is a plan view of the processing system illustrated inFIG. 1A, in accordance with one embodiment of the present invention;
• FIG. 1C is a side view illustrating a first processing rack in accordance with one embodiment of the present invention;
• FIG. 1D is a side view illustrating a second processing rack in accordance with one embodiment of the present invention;
• FIG. 1E is a plan view illustrating the processing system inFIG. 1B, in accordance with one embodiment of the present invention;
• FIG. 1F is a flow diagram illustrating a process sequence containing various process recipe steps that may be used in conjunction with the various embodiments of the cluster tool described herein;
• FIG. 1G is a plan view of the processing system illustrated inFIG. 1B that illustrates a transfer path of a substrate through the cluster tool following the process sequence illustrated inFIG. 1F;
• FIG. 1H is a block diagram illustrating a client server view of a computing environment and database system in accordance with one embodiment of the present invention;
• FIG. 2A schematically illustrates an example of a processing sequence according to one embodiment of the invention;
• FIG. 2B illustrates a timing diagram according to one embodiment of the invention;
• FIG. 2C illustrates a timing diagram according to one embodiment of the invention;
• FIG. 2D illustrates a series of timing diagrams according to one embodiment of the invention;
• FIG. 2E illustrates a series of timing diagrams according to one embodiment of the invention;
• FIG. 3A is a master move table for a three robot system in accordance with one embodiment of the present invention;
• FIG. 3B is another master move table for a three robot system in accordance with one embodiment of the present invention;
• FIG. 4 is a process flow diagram describing an offline method used to create a schedule for a track system in accordance with one embodiment of the present invention;
• FIG. 5 is a process flow diagram describing the pre-planning method used to create a schedule for a track system in accordance with one embodiment of the present invention;
• FIG. 6 is a process flow diagram describing the method used to create a schedule for a track system in accordance with one embodiment of the present invention;
• FIG. 7 is a simplified version of the cluster tool ofFIG. 1 illustrating different chamber zones in accordance with one embodiment of the present invention;
• FIGS. 8A-8C are robot move tables for thefirst robot assembly11A, thesecond robot assembly11B, and thethird robot assembly11C respectively;
• FIG. 9 is another master move table for a three robot system in accordance with one embodiment of the present invention;
• FIG. 10A is a chart illustrating the processing time for three lots of 25 wafers each using techniques known in the prior art; and
• FIG. 10B is a chart illustrating the processing time for three lots of 25 wafers each in accordance with one embodiment of the present invention.
• To facilitate understanding, identical reference numerals have been used, wherever possible, to designate identical elements that are common to the figures. It is contemplated that elements and/or process steps of one embodiment may be beneficially incorporated in other embodiments without additional recitation.
DETAILED DESCRIPTION
• Embodiments of the invention as recited in the claims generally provide methods and apparatus for processing substrates using a multi-chamber processing system (e.g., a cluster tool) that has an increased system throughput and repeatable wafer processing history.
• As used herein, the term “wafer history” generally refers to assuring that each substrate experiences the same repeatable substrate process timing, consistent exposure to process environment, and the timing between the various processing steps is the same for each substrate. A substrate's wafer history is generally monitored and controlled by process engineers to assure that all of the device fabrication processing variables that may later affect a device's performance are controlled, so that all substrates in the same batch are always processed the same way.
• As used herein, the terms “beat frequency” or “beat period” generally refer to a type of scheduling logic in which the robots in a cluster tool transfer and/or position substrates within set time intervals. Referring to the three robot system described herein, all three robots complete their transfer moves within the same time period, or beat. For example, movement between different chamber positions and pick/drop movements are done within the set interval or “beat.” Each “beat” may vary in length. The sum of all “beats,” or “beat period” defines the overall throughput of the tool of module for a given set of conditions. The logic used to define a series of “beats,” or “beat frequency” may be applied to systems containing two or more robots.
• As used herein, the term “cascading” generally refers to processing two or more lots or batches of wafers in a track system at the same time, for example, lot A and lot B. The goal of cascading is to minimize the gaps between lot A and lot B and thus increase substrate throughput, while maintaining a constant wafer history for the wafers in each lot. Minimizing the gaps between lot A and lot B increases the productivity of the cluster tool.
• As used herein, the term “substrate” generally refers to any workpiece upon which film processing is performed and may be used to denote a substrate, such as a semiconductor substrate or a glass substrate, as well as other material layers formed on the substrate, such as a dielectric layer.
• As used herein, the term “slack time” generally refers to “extra time.” For example, if a schedule requires a task to be completed in 102 seconds and the task takes 70 seconds, the slack time is 32 seconds. The process could be delayed to start up to 32 seconds, could start immediately and wait at the end for up to 32 seconds or split the time at the start and the end.
• To facilitate understanding, the following description refers to methods and apparatus using a three robot track system to perform a lithography process. However those skilled in the art will recognize that these methods and apparatus may be used with a number of different systems and processes.
• A. System Configuration
• FIG. 1A is an isometric view illustrating acluster tool10 in accordance with one embodiment of the present invention.FIG. 1A illustrates an embodiment of thecluster tool10 which contains three robot assemblies that are adapted to access the various process chambers that are stacked vertically in afirst processing rack60 and asecond processing rack80 and anexternal module5. In one aspect, when thecluster tool10 is used to complete a photolithography processing sequence theexternal module5, may be a stepper/scanner tool, that is attached to therear region45 to perform some additional exposure type processing step(s). One embodiment of thecluster tool10, as illustrated inFIG. 1A, contains afront end module24 and acentral module25.
• FIG. 1B is a plan view of the embodiment of thecluster tool10 shown inFIG. 1A, in accordance with one embodiment of the present invention. Thefront end module24 generally contains one or more pod assemblies105 (e.g.,items105A-D) and a front end robot assembly15 (FIG. 1B). The one or more pod assemblies105, or front-end opening unified pods (FOUPs), are generally adapted to accept one ormore cassettes106 that may contain one or more substrates “W”, or wafers, that are to be processed in thecluster tool10. In one aspect, thefront end module24 also contains one or more pass-through positions (e.g.,elements9A-CFIG. 1B).
• In one aspect, thecentral module25 has afirst robot assembly11A, asecond robot assembly11B, athird robot assembly11C, arear robot assembly40, afirst processing rack60 and asecond processing rack80. Thefirst processing rack60 and asecond processing rack80 contain various processing chambers (e.g., coater/developer chamber, bake chamber, chill chamber, wet clean chambers, etc. which are discussed below (FIGS.1C-D)) that are adapted to perform the various processing steps found in a substrate processing sequence.
• FIGS. 1C and 1D illustrate side views of one embodiment of thefirst processing rack60 andsecond processing rack80 as viewed when facing thefirst processing rack60 andsecond processing rack80 while standing on the side closest to side60A. Thefirst processing rack60 andsecond processing rack80 generally contain one or more groups of vertically stacked processing chambers that are adapted to perform some desired semiconductor or flat panel display device fabrication on a substrate. For example, inFIG. 1C thefirst process rack60 has five groups, or columns, of vertically stacked processing chambers. In general these device fabrication processes may include depositing a material on a surface of the substrate, cleaning a surface of the substrate, etching a surface of the substrate, or exposing the substrate to some form of radiation to cause a physical or chemical change to one or more regions on the substrate.
• In certain embodiments, thefirst processing rack60 andsecond processing rack80 have one or more processing chambers contained in them that can be adapted to perform one or more photolithography processing sequence steps. In one aspect, processing racks60 and80 may contain one or more coater/developer chambers160, one or morechill chambers180, one ormore bake chambers190, one or more optical edge bead removal (OEBR)chambers162, one or more post exposure bake (PEB)chambers130, one ormore support chambers165, an integrated bake/chill chamber (not shown), and/or one or more hexamethyldisilazane (HMDS) processingchambers170. Exemplary coater/developer chambers, chill chambers, bake chambers, OEBR chambers, PEB chambers, support chambers, integrated bake/chill chambers and/or HMDS processing chambers that may be adapted to benefit one or more aspects of the invention are further described in the commonly assigned U.S. patent application Ser. No. 11/112,281, filed Apr. 22, 2005, published as US 2006-0130750, which is hereby incorporated by reference in its entirety to the extent not inconsistent with the claimed invention. Examples of an integrated bake/chill chamber that may be adapted to benefit one or more aspects of the invention are further described in the commonly assigned U.S. patent application Ser. No. 11/111,154, filed Apr. 11, 2005, published as US 2006-0134340, and U.S. patent application Ser. No. 11/111,353, filed Apr. 11, 2005, published as US 2006-0130747, which are hereby incorporated by reference in their entirety to the extent not inconsistent with the claimed invention.
• Examples of processing chambers and or systems that may be adapted to perform one or more cleaning processes on a substrate and may be adapted to benefit one or more aspects of the invention is further described in the commonly assigned U.S. patent application Ser. No. 09/891,849, filed Jun. 25, 2001, published as US 2002-0029788, and U.S. patent application Ser. No. 09/945,454, filed Aug. 31, 2001, published as US 2003-0045098, which are hereby incorporated by reference in their entirety to the extent not inconsistent with the claimed invention.
• In one embodiment, as shown inFIG. 1C, where thecluster tool10 is adapted to perform a photolithography type process, thefirst processing rack60 may have eight coater/developer chambers160 (labeled CD1-8), eighteen chill chambers180 (labeled C1-18), eight bake chambers190 (labeled B1-8), six PEB chambers130 (labeled PEB1-6), two OEBR chambers162 (labeled162) and/or six HMDS process chambers170 (labeled DP1-6). In one embodiment, as shown inFIG. 1D, where thecluster tool10 is adapted to perform a photolithography type process, thesecond process rack80 may have eight coater/developer chambers160 (labeled CD1-8), six integrated bake/chill chambers800 (labeled BC1-6), six HMDS process chambers170 (labeled DP1-6) and/or six support chambers165 (labeled S1-6). The orientation, positioning, type, and number of process chambers shown in the FIGS.1C-D are not intended to be limiting as to the scope of the invention, but are intended to illustrate an embodiment of the invention.
• Referring toFIG. 1B, in one embodiment, the frontend robot assembly15 is adapted to transfer substrates between acassette106 mounted in a pod assembly105 (seeelements105A-D) and the one or more of the pass-through positions (see pass-throughpositions9A-C inFIG. 1B). In another embodiment, the frontend robot assembly15 is adapted to transfer substrates between a cassette mounted in a pod assembly105 and the one or more processing chambers in the first processing racks60 or asecond processing rack80 that abuts thefront end module24. The frontend robot assembly15 generally contains ahorizontal motion assembly15A and arobot15B, which in combination are able to position a substrate in a desired horizontal and/or vertical position in thefront end module24 or the adjoining positions in thecentral module25. The frontend robot assembly15 is adapted to transfer one or more substrates using one or more robot blades15C, by use commands sent from a system controller101 (discussed below). In one sequence the frontend robot assembly15 is adapted to transfer a substrate from thecassette106 to one of the pass-through positions (e.g.,elements9A-C inFIG. 1B). Generally, a pass-through position is a substrate staging area that may contain a pass-through processing chamber that has features similar to anexchange chamber533, or aconventional substrate cassette106, and is able to accept a substrate from a first robot assembly so that it can be removed and repositioned by a second robot assembly. In one aspect, the pass-through processing chamber mounted in a pass-through position may be adapted to perform one or more processing steps in a desired processing sequence, for example, a HMDS process step or a chill/cooldown processing step or substrate notch align. In one aspect, each of the pass-through positions (elements9A-C inFIG. 1B) may be accessed by each of the central robot assemblies (i.e.,first robot assembly11A,second robot assembly11B, andthird robot assembly11C).
• Referring to FIGS.1A-B, thefirst robot assembly11A, thesecond robot assembly11B, and thethird robot assembly11C are adapted to transfer substrates to the various processing chambers contained in thefirst processing rack60 and thesecond processing rack80. In one embodiment, to perform the process of transferring substrates in thecluster tool10 thefirst robot assembly11A, thesecond robot assembly11B, and thethird robot assembly11C have similarly configuredrobot assemblies11 which each have at least onehorizontal motion assembly90, avertical motion assembly95, and arobot hardware assembly85 which are in communication with asystem controller101. In one aspect, theside60B of thefirst processing rack60, and theside80A of thesecond processing rack80 are both aligned along a direction parallel to the horizontal motion assembly90 (described below) of each of the various robot assemblies (i.e.,first robot assembly11A,second robot assembly11B,third robot assembly11C).
• Thesystem controller101 is adapted to control the position and motion of the various components used to complete the transferring process. Thesystem controller101 is generally designed to facilitate the control and automation of the overall system and typically includes a central processing unit (CPU) (not shown), memory (not shown), and support circuits (or I/O) (not shown). The CPU may be one of any form of computer processors that are used in industrial settings for controlling various system functions, chamber processes and support hardware (e.g., detectors, robots, motors, gas sources hardware, etc.) and monitor the system and chamber processes (e.g., chamber temperature, process sequence throughput, chamber process time, I/O signals, etc.). The memory is connected to the CPU, and may be one or more of a readily available memory, such as random access memory (RAM), read only memory (ROM), floppy disk, hard disk, or any other form of digital storage, local or remote. Software instructions and data can be coded and stored within the memory for instructing the CPU. The support circuits are also connected to the CPU for supporting the processor in a conventional manner. The support circuits may include cache, power supplies, clock circuits, input/output circuitry, subsystems, and the like. A program (or computer instructions) readable by thesystem controller101 determines which tasks are performable on a substrate. Preferably, the program is software readable by thesystem controller101 that includes code to perform tasks relating to monitoring and execution of the processing sequence tasks and various chamber process recipe steps.
• Referring toFIG. 1B, in one aspect of the invention thefirst robot assembly11A is adapted to access and transfer substrates between the processing chambers in thefirst processing rack60 from at least one side, e.g., theside60B. In one aspect, thethird robot assembly11C is adapted to access and transfer substrates between the processing chambers in thesecond processing rack80 from at least one side, e.g., theside80A. In one aspect, thesecond robot assembly11B is adapted to access and transfer substrates between the processing chambers in thefirst processing rack60 fromside60B and thesecond processing rack80 fromside80A.FIG. 1E illustrates a plan view of the embodiment of thecluster tool10 shown inFIG. 1B, in which arobot blade87 from thesecond robot assembly11B has been extended into a processing chamber in thefirst processing rack60 throughside60B. The ability to extend therobot blade87 into a processing chamber and retract therobot blade87 from the processing chamber is generally completed by cooperative movement of the components contained in thehorizontal motion assembly90,vertical motion assembly95, androbot hardware assembly85, and by use of commands sent from thesystem controller101. The ability of two or more robots to “overlap” with one another, such as thefirst robot assembly11A and thesecond robot assembly11B or thesecond robot assembly11B and thethird robot assembly11C, is advantageous since it allows substrate transfer redundancy which can improve the cluster reliability, uptime, and also increase the substrate throughput. Robot “overlap” is generally the ability of two or more robots to access and/or independently transfer substrates between the same processing chambers in the processing rack. The ability of two or more robots to redundantly access processing chambers can be an important aspect in preventing system robot transfer bottlenecks, since it allows an under utilized robot to help out a robot that is limiting the system throughput. Therefore, the substrate throughput can be increased, a substrate's wafer history can be made more repeatable, and the system reliability can be improved through the act of balancing the load that each robot takes during the processing sequence.
• In one aspect of the invention, the various overlapping robot assemblies (e.g.,elements11A,11B,11C,11D,11E, etc.) are able to simultaneously access processing chambers that are horizontally adjacent (x-direction) or vertically adjacent (z-direction) to each other. For example, when using the cluster tool configurations illustrated inFIGS. 1B and 1C, thefirst robot assembly11A is able to access processing chamber CD6 in thefirst processing rack60 and thesecond robot assembly11B is able to access processing chamber CD5 simultaneously without colliding or interfering with each other. In another example, when using the cluster tool configurations illustrated inFIGS. 1B and 1D, thethird robot assembly11C is able to access processing chamber C6 in thesecond processing rack80 and thesecond robot assembly11B is able to access processing chamber P6 simultaneously without colliding or interfering with each other.
• In one aspect, thesystem controller101 is adapted to adjust the substrate transfer sequence through the cluster tool based on a calculated optimized throughput or to work around processing chambers that have become inoperable. A feature of thesystem controller101 which enhances throughput is known as the logical scheduler. The logical scheduler prioritizes tasks and substrate movements based on inputs from the user and various sensors distributed throughout the cluster tool. The logical scheduler may be adapted to review the list of future tasks requested of each of the various robots (e.g., frontend robot assembly15,first robot assembly11A,second robot assembly11B,third robot assembly11C, etc.), which are retained in the memory of thesystem controller101, to help balance the load placed on each of the various robots. The use of asystem controller101 to maximize the utilization of the cluster tool will improve the cluster tool's CoO, makes the wafer history more repeatable, and can improve the cluster tool's reliability.
• In one aspect, thesystem controller101 is also adapted to prevent collisions between the various overlapping robots and increase the substrate throughput. In one aspect, thesystem controller101 is further programmed to monitor and control the motion of thehorizontal motion assembly90, avertical motion assembly95, and arobot hardware assembly85 of all the robots in the cluster tool to avoid a collision between the robots and improve system throughput by allowing all of the robots to be in motion at the same time. This so called “collision avoidance system,” may be implemented in multiple ways, but in general thesystem controller101 monitors the position of each of the robots by use of various sensors positioned on the robot(s) or in the cluster tool during the transferring process to avoid a collision. In one aspect, the system controller is adapted to actively alter the motion and/or trajectory of each of the robots during the transferring process to avoid a collision and minimize the transfer path length. Examples of other cluster tool systems are disclosed in commonly assigned U.S. patent application Ser. No. 11/315,984, filed Dec. 22, 2005, published as US 2006-0182536, which is hereby incorporated by reference in its entirety to the extent not inconsistent with the claimed invention.
• B. Transfer Sequence Example
• FIG. 1F illustrates one example of asubstrate processing sequence500 through thecluster tool10, where a number of process steps (e.g., elements501-520) may be performed after each of the transferring steps A₁-A₁₀have been completed. One or more of the process steps501-520 may entail performing vacuum and/or fluid processing steps on a substrate, to deposit a material on a surface of the substrate, to clean a surface of the substrate, to etch a surface of the substrate, or to exposing the substrate to some form of radiation to cause a physical or chemical change to one or more regions on the substrate. Examples of typical processes that may be performed are photolithography processing steps, substrate clean process steps, CVD deposition steps, ALD deposition steps, electroplating process steps, or electroless plating process steps.
• Thesubstrate processing sequence500 may be a photolithographic process. The photolithographic process may generally contain the following: a remove substrate from pod step, a BARC coat step, a post BARC bake step, a post BARC chill step, a photoresist coat step, a post photoresist coat bake step, a post photoresist chill step, an optical edge bead removal (OEBR) step, an exposure step, a post exposure bake (PEB) step, a post PEB chill step, a develop step, and a place in pod step. In other embodiments, the sequence of the method steps500 may be rearranged, altered, one or more steps may be removed, or two or more steps may be combined into a single step without varying from the basic scope of the invention.
• The remove substrate from pod step is generally defined as the process of having thefront end robot15 remove a substrate from acassette106 resting in one of the pod assemblies105. Acassette106, containing one or more substrates “W”, is placed on the pod assembly105 by the user or some external device (not shown) so that the substrates can be processed in thecluster tool10 by a user-defined substrate processing sequence controlled by software retained in thesystem controller101.
• The BARC coat step, or bottom anti-reflective coating process (hereafter BARC), is used to deposit an organic material over a surface of the substrate. The BARC layer is typically an organic coating that is applied onto the substrate prior to the photoresist layer to absorb light that otherwise would be reflected from the surface of the substrate back into the photoresist during the exposure step performed in the stepper/scanner. If these reflections are not prevented, optical standing waves will be established in the photoresist layer, which cause feature size(s) to vary from one location to another depending on the local thickness of the photoresist layer. The BARC layer may also be used to level (or planarize) the substrate surface topography, since surface topography variations are invariably present after completing multiple electronic device fabrication steps. The BARC material fills around and over the features to create a flatter surface for photoresist application and reduces local variations in photoresist thickness. The BARC coat step is typically performed using a conventional spin-on photoresist dispense process in which an amount of the BARC material is deposited on the surface of the substrate while the substrate is being rotated, which causes a solvent in the BARC material to evaporate and thus causes the material properties of the deposited BARC material to change. The air flow and exhaust flow rate in the BARC processing chamber is often controlled to control the solvent vaporization process and the properties of the layer formed on the substrate surface.
• The post BARC bake step is used to assure that all of the solvent is removed from the deposited BARC layer in the BARC coat step, and in some cases to promote adhesion of the BARC layer to the surface of the substrate. The temperature of the post BARC bake step is dependent on the type of BARC material deposited on the surface of the substrate, but will generally be less than about 250° C. The time required to complete the post BARC bake step will depend on the temperature of the substrate during the post BARC bake step, but will generally be less than about 60 seconds.
• The post BARC chill step, is used to assure that the time the substrate is at a temperature above ambient temperature is controlled so that every substrate sees the same time-temperature profile; thus process variability is minimized. Variations in the BARC process time-temperature profile, which is a component of a substrate's wafer history, can have an effect on the properties of the deposited film layer and thus is often controlled to minimize process variability. The post BARC chill step is typically used to cool the substrate after the post BARC bake step to a temperature at or near ambient temperature. The time required to complete the post BARC chill step will depend on the temperature of the substrate exiting the post BARC bake step, but will generally be less than about 30 seconds.
• The photoresist coat step is used to deposit a photoresist layer over a surface of the substrate. The photoresist layer deposited during the photoresist coat step is typically a light sensitive organic coating that is applied onto the substrate and is later exposed in the stepper/scanner to form the patterned features on the surface of the substrate. The photoresist coat step is typically performed using conventional spin-on photoresist dispense process in which an amount of the photoresist material is deposited on the surface of the substrate while the substrate is being rotated, thus causing a solvent in the photoresist material to evaporate and the material properties of the deposited photoresist layer to change. The air flow and exhaust flow rate in the photoresist processing chamber is controlled to control the solvent vaporization process and the properties of the layer formed on the substrate surface. In some cases it may be necessary to control the partial pressure of the solvent over the substrate surface to control the vaporization of the solvent from the photoresist during the photoresist coat step by controlling the exhaust flow rate and/or by injecting a solvent near the substrate surface. To complete the photoresist coat step the substrate is first positioned on a spin chuck in a coater chamber. A motor rotates the spin chuck and substrate while the photoresist is dispensed onto the center of the substrate. The rotation imparts an angular torque onto the photoresist, which forces the photoresist out in a radial direction, ultimately covering the substrate.
• The post photoresist coat bake step is used to assure that most, if not all, of the solvent is removed from the deposited photoresist layer in the photoresist coat step, and in some cases to promote adhesion of the photoresist layer to the BARC layer. The temperature of the post photoresist coat bake step is dependent on the type of photoresist material deposited on the surface of the substrate, but will generally be less than about 250° C. The time required to complete the post photoresist coat bake step will depend on the temperature of the substrate during the post photoresist bake step, but will generally be less than about 60 seconds.
• The post photoresist chill step is used to control the time the substrate is at a temperature above ambient temperature so that every substrate sees the same time-temperature profile and thus process variability is minimized. Variations in the time-temperature profile can have an affect on properties of the deposited film layer and thus is often controlled to minimize process variability. The temperature of the post photoresist chill step is thus used to cool the substrate after the post photoresist coat bake step to a temperature at or near ambient temperature. The time required to complete the post photoresist chill step will depend on the temperature of the substrate exiting the post photoresist bake step, but will generally be less than about 30 seconds.
• The optical edge bead removal (OEBR) step, is a process used to expose the deposited light sensitive photoresist layer(s), such as the layers formed during the photoresist coat step and the BARC layer formed during the BARC coat step, to a radiation source (not shown) so that either or both layers can be removed from the edge of the substrate and the edge exclusion of the deposited layers can be more uniformly controlled. The wavelength and intensity of the radiation used to expose the surface of the substrate will depend on the type of BARC and photoresist layers deposited on the surface of the substrate. An OEBR tool can be purchased, for example, from USHIO America, Inc. Cypress, Calif.
• The exposure step is a lithographic projection step applied by a lithographic projection apparatus (e.g., stepper scanner) to form a pattern which is used to manufacture integrated circuits (ICs). The exposure step forms a circuit pattern corresponding to an individual layer of the integrated circuit (IC) device on the substrate surface, by exposing the photosensitive materials, such as, the photoresist layer formed during the photoresist coat step and the BARC layer formed during the BARC coat step (photoresist) of some form of electromagnetic radiation. The stepper/scanner may be purchased from Cannon, Nikon, or ASML.
• The post exposure bake (PEB) step is used to heat a substrate immediately after the exposure step in order to stimulate diffusion of the photoactive compound(s) and reduce the effects of standing waves in the photoresist layer. For a chemically amplified photoresist, the PEB step also causes a catalyzed chemical reaction that changes the solubility of the photoresist. The control of the temperature during the PEB is critical to critical dimension (CD) control. The temperature of the PEB step is dependent on the type of photoresist material deposited on the surface of the substrate, but will generally be less than about 250° C. The time required to complete the PEB step will depend on the temperature of the substrate during the PEB step, but will generally be less than about 60 seconds.
• The post exposure bake (PEB) chill step is used to assure that the time the substrate is at a temperature above ambient temperature is controlled, so that every substrate sees the same time-temperature profile and thus process variability is minimized. Variation in the PEB process time-temperature profile can have an effect on properties of the deposited film layer and thus is often controlled to minimize process variability. The temperature of the post PEB chill step is thus used to cool the substrate after the PEB step to a temperature at or near ambient temperature. The time required to complete the post PEB chill step will depend on the temperature of the substrate exiting the PEB step, but will generally be less than about 30 seconds.
• The develop step is a process in which a solvent is used to cause a chemical or physical change to the exposed or unexposed photoresist and BARC layers to expose the pattern formed during the exposure step. The develop process may be a spray or immersion or puddle type process that is used to dispense the developer solvent. In one embodiment of the develop step, after the solvent has been dispensed on the surface of the substrate a rinse step may be performed to rinse the solvent material from the surface of the substrate. The rinse solution dispensed on the surface of the substrate may contain deionized water and/or a surfactant.
• The insert the substrate in pod step is generally defined as the process of having thefront end robot15 return the substrate to acassette106 resting in one of the pod assemblies105.
• FIG. 1G illustrates an example of the transfer steps that a substrate may follow as it is transferred through a cluster tool that is configured as the cluster tool shown inFIG. 1B following theprocessing sequence500 described inFIG. 1F. In this embodiment, the substrate is removed from a pod assembly105 (item #105D) by the frontend robot assembly15 and is delivered to a chamber positioned at the pass-throughposition9C following the transfer path A₁, so that the pass-throughstep502 can be completed on the substrate. In one embodiment, the pass-throughstep502 entails positioning or retaining the substrate so that another robot could pickup the substrate from the pass-throughposition9C. Once the pass-throughstep502 has been completed, the substrate is then transferred to afirst process chamber531 by thethird robot assembly11C following the transfer path A₂, whereprocess step504 is completed on the substrate. After completing theprocess step504 the substrate is then transferred to thesecond process chamber532 by thethird robot assembly11C following the transfer path A₃. After performing theprocess step506 the substrate is then transferred by thesecond robot assembly11B, following the transfer path A₄, to theexchange chamber533. After performing theprocess step508 the substrate is then transferred by therear robot assembly40, following the transfer path A₅, to theexternal processing system536 where theprocess step510 is performed. After performingprocess step510 the substrate is then transferred by arear robot assembly40, following the transfer path A₆, to theexchange chamber533 where theprocess step512 is performed. In one embodiment, the process steps508 and512 entail positioning or retaining the substrate so that another robot could pickup the substrate from theexchange chamber533. After performing theprocess step512 the substrate is then transferred by thesecond robot assembly11B, following the transfer path A₇, to theprocess chamber534 where theprocess step514 is performed. The substrate is then transferred to processchamber535 following the transfer path A₈using thefirst robot assembly11A. After theprocess step516 is complete, thefirst robot assembly11A transfers the substrate to a pass-through chamber positioned at the pass-throughposition9A following the transfer path A₉. In one embodiment, the pass-throughstep518 entails positioning or retaining the substrate so that another robot could pickup the substrate from the pass-throughposition9A. After performing the pass-throughstep518 the substrate is then transferred by the frontend robot assembly15, following the transfer path A₁₀, to thepod assembly105D.
• In one embodiment, process steps504,506,510,514, and516 are a photoresist coat step, a bake/chill step, an exposure step performed in a stepper/scanner module, a post exposure bake/chill step, and a develop step, respectively, which are further described in the commonly assigned U.S. patent application Ser. No. 11/112,281, filed Apr. 22, 2005, published as 2006-0130750, which is incorporated by reference herein. The bake/chill step and the post exposure bake/chill steps may be performed in a single process chamber or they may also be transferred between a bake section and a chill section of an integrated bake/chill chamber by use of an internal robot (not shown). WhileFIGS. 1F-1G illustrate one example of a process sequence that may be used to process a substrate in acluster tool10, process sequences and/or transfer sequences that are more or less complex may be performed without varying from the basic scope of the invention.
• Also, in one embodiment, thecluster tool10 is not connected to or in communication with anexternal processing system536 and thus therear robot assembly40 is not part of the cluster tool configuration and the transfer steps A₅-A₆andprocess step510 are not performed on the substrate. In this configuration all of the processing steps and transferring steps are performed between positions or processing chambers within in thecluster tool10.
• FIG. 1H is a block diagram that illustrates a client server view of acomputing environment600, includingcontroller101, andcluster tool10, according to one embodiment of the invention. As shown,computing environment600 includes aclient computer system610,network615,server system620, andcontroller101. In one embodiment, the computer systems illustrated inenvironment600 may include computer existing computer systems, e.g., desktop computers, server computers laptop computers, tablet computers, and the like. The software applications described herein, however, are not limited to any particular computing system or application or network architecture and may be adapted to take advantage of new computing systems as they become available. Additionally, those skilled in the art will recognize that the computer systems shown inFIG. 1H are simplified to highlight aspects of the present invention and that computing systems and networks typically include a variety of additional elements not shown inFIG. 1H.
• As shown,client computer system610 includes aCPU602,storage604 andmemory606, typically connected by a bus (not shown).CPU602 is a programmable logic device that performs all the instruction, logic, and mathematical processing in a computer.Storage604 stores application programs and data for use by aclient computer system610.Storage604 includes hard-disk drives, flash memory devices, optical media and the like.Network615 generally represents any kind of data communications network. Accordingly,network615 may represent both local and wide area networks, including the Internet.
• Server620 also includes aCPU622,storage624 andmemory626. As shown, seversystem620 also includes a database management system (DBMS)630 in communication withdatabase640. TheDBMS630 includes software used to organize, analyze, and modify information stored in adatabase640.Database640 contains the data managed byDBMS630. At various times elements ofdatabase640 may be present instorage624 andmemory626.
• C. Wafer Transfer System
• Beat Frequency Control
• As noted above one of the key elements to a desirable process sequence performed on a substrate in a cluster tool system is the need to assure that each wafer processed has a similar wafer history. Typical conventional processing tools use algorithms or software that is often called “adaptive sequencers” to control the movement of substrates through the cluster tool. Decisions made by the adaptive sequencers are made based on demands or events that normally occur during the processing and/or during fault situations. In this configuration, to plan or control the current and future movement of wafers through the cluster tool, the system controller makes decisions based on a defined and prioritized set of rules that are applied to a situation based on the specific faults or specific events that may be occurring at that instant in time. It is thus easy to see that different actions may be taken at any given instant in time based on various process sequence variables, such as the position of the wafers in the system or the faults that have arisen during the processing of the substrate. This demand based control will cause or create a variation in the movement of wafers and the overall control of the system. Adaptive sequencers will also require a system to contain and use logic components that are able to make decisions and then control and carryout these decisions accordingly, which can be costly. Use of adaptive type sequencers will affect the wafer history experienced by each substrate and decrease wafer throughput time through the manufacturing equipment, since the timing between processing or transferring steps may vary due to task conflicts created by the changing priorities made by the adaptive sequencers.
• Therefore, there is a need for a system, a method and an apparatus that can process a substrate so that it can meet the required device performance goals and increase the system throughput while maintaining constant “wafer history” and reducing the system cost and complexity. To resolve these issues the following control architecture, techniques, algorithms, and systems are used to control the substrate throughput through a cluster tool. These elements may include beat frequency control, an offline application, and wafer cascading type control, which are discussed herein.
• One aspect of the control system discussed herein is the use of software and modeling techniques that allow a user to optimize and define an optimal processing sequence based on the user's goals. In general, the user defined processing sequence will be maintained or proportionally scaled so that the key goals of the user define processing sequence will not change. This is generally performed by use of an offline application that requires the user to make decisions as to what features are important to achieve a desired substrate processing sequence. These decisions may include the requirement that the wafer history for each wafer be as uniform as possible, that the throughput be as high as possible, that the wafers not wait at certain points in the processing sequence, and/or that wafers going into the system only see one robot blade while wafers leaving the system see another robot blade, to name just a few. The offline application is discussed further below.
• One aspect of the process sequence control is a wafer throughput management process called beat frequency. The “beat frequency” or “beat period” refers to a type of scheduling logic in which all robots within a cluster tool transfer and/or position substrates within a series of set time intervals. This scheduling logic assures that the robot to robot interaction and substrate movement remains constant and well known since the movement or actions taken by the robots are started at the same time and are not completed until all the robots have performed the defined task within the set time interval. For example, referring to the three robot system described herein, all three robots move at the same time to a different chamber position and then each robot performs a Pick/Drop movement at the same time. Each of these actions/motions is done within a set interval or “beat.” Each “beat” may vary in length. The sum of all “beats,” or “beat period” defines the overall throughput of the tool or module for a given set of conditions. The logic used to define a series of “beats,” or “beat frequency” may be applied to systems containing two or more robots.
• FIG. 2A schematically illustrates an example of aprocessing sequence1001 that a plurality of substrates follow during processing in a three robot containing cluster tool. The cluster tool has a first robot that is adapted to service a first group ofchambers1010, a second robot that is adapted to service a second group ofchambers1020 and a third robot that is adapted to service a third group ofchambers1030. The first robot is adapted to service four chambers in the first group ofchambers1010, which include the cassette CA1 and process chambers A-C. The second robot is adapted to service three chambers in the second group ofchambers1020, which include the process chambers C-E. The third robot is adapted to service four chambers in the third group ofchambers1030, which include the process chambers E-G and cassette CA1. As shown inFIG. 2A, each of the robots in the system will overlap at least one point in the system, such as the first robot and the second overlap at chamber C, and the second and third robots overlap at chamber E, thus providing places where the robots can collide or transferring bottlenecks can occur. The configuration of the chambers, the number of chambers in each group and the number of robots in the cluster in this example are not intended to be limiting as to the scope of the invention discussed herein.
• To transfer substrates through the system the first robot performs threemovement steps1005A-1005C, the second robot performs two movement steps1006A-1006B, and the third robot performs threemovement steps1007A-1007C. Each of the movement steps1005A-1005C,1006A-1006B, and1007A-1007C performed by each robot will require two sub-steps in which 1) the robot physically moves from one position to another within the cluster tool, which are referred to hereafter as “moves”, and 2) an extension step is performed in which the substrate is positioned by use of a robot blade type of device. Typically, the extension steps only require the robot blade type to be translated to a desired position within the cluster tool (e.g., into a processing chamber, into a buffer station), while the major components (e.g.horizontal assembly90,vertical assembly95 inFIG. 1B) in the robot are generally stationary. Extension steps are actions in which the system controller causes the robot blade to extend or retract to cause the wafer to be placed or removed from a desired position within the cluster tool. The wafer placement or removal steps are some times called hereafter a “drop” or a “pick” move, respectively. In some cases it may be desirable to overlap a portion of the move and extension steps to improve substrate throughput, and thus the illustration of completely isolated steps inFIGS. 2A-2E are not intended to be limiting to the scope of the invention described herein.
• FIG. 2B is a schematic view of a timing diagram1040 that illustrates a series of actions performed by a robot in which the robot performs a number of moves (e.g., M₁, M₂, M₃) that are followed by a number of extension steps (e.g., T₁, T₂, T₃). Each of the transfer steps1041-1046 contain a move (e.g., M₁-M₆) and an extension step (e.g., T₁-T₆). In one example, as shown inFIG. 2B each move (e.g., M₁, M₂, M₃) has a similar length, and each extension step (e.g., T₁, T₂, T₃) has an equivalent length. In most real systems the move times may vary due to the physical spacing that the robot is required to translate during each move, while the extension steps will generally remain fairly constant, since these movements are limited by the need to make sure the wafer does not slide or move on the blade during the translation of the substrate. The timing diagram1040 is generally intended to illustrate an example of a transferring sequence performed by a single bladed robot assembly, since each of the transfer steps require moves (e.g., M₁-M₆) that have a similar duration in time.
• FIG. 2C is a schematic view of a timing diagram1050 that illustrates a series of transfer steps performed by a robot in which the moves (e.g., M₁, M₂, M₃) have differing lengths and the extension steps (e.g., T₁, T₂, T₃) have a fairly similar duration. This example is similar to a transfer process that may be performed by a dual blade assembly used to swap and transfer wafers from various positions within the cluster tool. In general, a dual blade assembly has two independently controlled robotic arms that are coupled together and remain in a known orientation to each other so that each of the two independently controlled robotic arms can independently position a wafer in two or more know positions in the cluster tool. In one embodiment, during each of the moves (e.g., M₁, M₂, M₃) both of the robot blades are transferred together from one position to another position within the cluster tool. An example of a dual blade robot assembly is further described in the co-pending and commonly assigned U.S. patent application Ser. No. 11/315,984, filed Dec. 22, 2005, entitled CARTESIAN ROBOT CLUSTER TOOL ARCHITECTURE, which is incorporated by reference herein.
• In one embodiment, as shown inFIG. 2C, during atransfer step1051 the dual bladed robot performs a move M₁and then a wafer is picked-up from a first extension position within the cluster tool by use of an extension step T₁performed by one of the robot's dual blades. Next the robot performs aswap transfer step1052 in which the dual bladed robot performs a move M₂, a second extension step T₂, a short move M₃, and then third extension step T₃. The second move M₂allows the wafer and the robotic arms to be moved from a first position to a second position so that the robotic arm that the non-wafer containing blade can remove (“pick” step) a wafer positioned in a second extension position during the second extension step T₂. The short move M₃is generally shorter than the other moves M₁and M₂, since the only action occurring in this step is a short translational movement that allows a robotic arm that contains a wafer to move into alignment with the second extension position from a position that was close to the second extension position, and so that a third extension step T₃can be performed. In this example, the robot then performs aswap transfer step1053 in which the dual bladed robot performs a move M₄to a fourth position within the cluster tool, a fourth extension step T₄to a third extension position, a short move M₅, and then fifth extension step T₅to the third extension position. The actual movement and actions taken in theswap step1053 are similar to theswap transfer step1052 discussed above, except for movement between differing positions within the cluster tool and thus are not discussed again here. Next, the wafer remaining on one of the blades after theswap transfer step1053 is dropped off in a desired position within the cluster tool during atransfer step1054 in which the dual bladed robot performs a move M₆and then the wafer is dropped off to a fourth extension position within the system by use of an extension step T₆. Although the number of steps performed during the timing diagram1040 (FIG. 2B) and the timing diagram1050 (FIG. 2C) is the same, the wafer throughput inFIG. 2C is higher due to the reduction in the time required to perform moves M₃and M₅.
• FIG. 2D illustrates a timing diagram1060 that illustrates a series of transfer steps performed by three robots in which the first robot completes moves (e.g., M₁₁, M₁₂, M₁₃, M₁₄) and extension steps (e.g., T₁₁, T₁₂, T₁₃, T₁₄), the second robot completes moves (e.g., M₂₁, M₂₂, M₂₃, M₂₄) and extension steps (e.g., T₂₁, T₂₂, T₂₃), and the third robot completes moves (e.g., M₃₁, M₃₂, M₃₃, M₃₄) and extension steps (e.g., T₃₁, T₃₂, T₃₃, T₃₄) that each may have differing lengths in time. As shown inFIG. 2D, the relationship of each of the robots to each other may vary at any instant in time and thus the rules or timing between processes may vary, thus causing the system controller to make decisions on-the-fly.
• FIG. 2E illustrates a series of timing diagrams1140,1150 and1160 for each of the three robots (e.g.,first robot assembly11A,second robot assembly11B,third robot assembly11C inFIG. 1B) used to perform theprocessing sequence1001 discussed inFIG. 2A, and is used herein to describe the various aspects of beat frequency type of control. The timing diagram1180 illustrates the timing of each beat that correlates to the move and transfers steps illustrated in the timing diagrams1140,1150 and1160. Thetime line1170 represents the duration of each of the beats in the timing diagram1180 and their relation to real time.
• In this configuration, the length of time required to complete each simultaneously performed move and/or extension step will occur during the same length of time. In general, this means that each move and/or extension step will start at about the same time and end at different times, but the robots that perform the shorter duration steps will remain idle for a period of time until the longest step performed by a robot is completed. For example, during the first beat, orBeat1, all three robots perform their respective moves (i.e., M₁₁, M₂₁, and M₃₁) during the first portion of the movement steps1005A,1006A and1007A, respectively. In this example, the first robot performs a move M₁₁, that takes 1.8 seconds, the second robot performs a move M₂₁that takes 1.5 seconds, and the third robot performs a move M₃₁that takes 1.7 seconds. Thus the length ofBeat1 is 1.8 seconds, and the next part of the movement steps1005A,1006A and1007A are not started until the first robot has finished its move M₁₁and theBeat1 has ended. Next, in this example, the first robot performs an extension step T₁₁that takes 2.0 seconds, the second robot performs an extension step T₂₁that takes 1.9 seconds and the third robot performs an extension step T₃₁that takes 2.1 seconds. Thus the length ofBeat2 is 2.1 seconds, and the movement steps1005B and1007B are not started until the third robot has finished its extension step T₃₁and theBeat2 has ended. The beats and sub-beats for all of the subsequent steps are similarly performed in the processing sequence so that each of the simultaneous performed steps are performed within a defined beat and the length of the beat is set by the length of the longest step. The total time to execute these 6 beats is thus approximately 11.7 seconds. As illustrated inFIG. 2E, in some cases an under-utilized robot, such as the second robot (second robot assembly11B) robot may remain idle for one or more of the beats to help prevent collisions between robots.
• In one aspect, it may be desirable to adjust the phase of the movement steps, or sub-steps, to prevent the collisions that can occur when two or more robots are required to pick or place a wafer in the same chamber, or at a different chamber positioned in the same vertical rack, during the same beat. In one embodiment, it may be required to continually alter the phase of the movement steps to prevent collisions from occurring at multiples of the beats, such as first robot and second robot will move into phase every sixth beat. In another embodiment, it may be necessary to adjust the phasing of the beats by shifting the collision prone cyclic occurring beats so that they are out of phase, and then adding one or more idle periods into a robots' transferring sequence to assure that each robot has the same number of beats in the cycle, or the various robots have a number of beats that are multiples of each other in the repeating cycle, so that the phase relationship never varies with time.
• In one embodiment, it may be desirable for one or more robots in a cluster tool to perform one or more “ghost moves” during a beat within a processing sequence to assure that the physical relationship of the robots to each other remains consistent during each part of the processing sequence. A “ghost move” refers to a movement of the robot in which the robot goes through the motion of transferring a wafer that is not present on the robot blade or in the processing chamber. The “ghost move” is thus used to maintain a similar relationship of two or more robots to each other to help make the wafer history more repeatable.
• While not illustrated inFIGS. 2A-2E for complexity reasons, the optimum beat frequency will generally also take into account the variations in processing times of each of the process chambers (e.g., chambers A-G inFIG. 2A) in theprocessing sequence1001 to assure that wafers are not left idle for varying amounts of time during the process sequence. Thus, in one embodiment, it is desirable to alter the length of one or more of the beats to assure that the timing to pickup and/or drop-off a substrate is more repeatable. It is believed that during amplified resist photolithography type processes this analysis can be important since it is not desirable to allow wafers exiting a stepper (e.g.,external module5 inFIGS. 1A-1E) to wait a variable amount of time due to the typical generation of PAGs that will affect CD of the processed substrate.
• In another embodiment, it is desirable to strategically add one or more buffer stations within a cluster tool to account for systematic variations in substrate throughput caused by the cyclic nature of the transfer sequences and the shifting bottlenecks in the system due to system faults or other typical unknown causes of throughput variability. The buffer stations generally contain a plurality of wafer holding positions (e.g., support chambers165) that may be accessible by one or more of the robots and allow substrates to collect so that the timing between critical steps can be maintained. In general the buffer positions will be placed in positions where waiting a variable amount of time will not affect the process results by varying the wafer history of each wafer, such as after a cool down step during photolithography processing. An example of various exemplary buffer stations that may be used in conjunction with the embodiments described herein are further described in the in the co-pending and commonly assigned U.S. patent application Ser. No. 11/315,984, filed Dec. 22, 2005, entitled CARTESIAN ROBOT CLUSTER TOOL ARCHITECTURE, which is incorporated by reference herein.
EXAMPLE
• FIGS. 3A-3B are robot move tables for thefirst robot assembly11A, thesecond robot assembly11B, and thethird robot assembly11C, respectively.FIG. 3A is an example of actual data collected for a robotic system, or hereafter a master move table1300, for a three robot system representing a desirable process flow that has a maximum robot limited throughput. The data represented in the “zone” columns of the tables shown inFIGS. 3A and 3B relate toFIG. 7, which is a schematic plan view of asimplified version1700 of thecluster tool10 ofFIG. 1 showingdifferent chamber zones1705,1710,1715,1720,1725, and1730 in which therobots11A,11B and11C may move. In this example, thechamber zones1705,1710,1715,1720,1725, and1730 are adjacent to the various processing chambers found in thefirst processing rack60 andsecond processing rack80. In general, all threerobots11A,11B, and11C move to different chamber positions within different chamber zones at the same time and will also initiate transfer moves (pick/drop) at the same time, this is referred to as the “beat frequency,” which is described above. A more complete discussion of a master move table is discussed below in conjunction withFIG. 6.
• FIGS. 3A-3B are robot move tables that represent groups of moves (M) and extension steps (i.e., D/P (or Drop/Pick) in the “Move Type” column) that are sequentially performed by therobots11A,11B and11C. Each row illustrates transferring process step(s) taken by each of the robots during each beat that lasted a defined time shown in the “time” column. The moves performed during the transferring process have been characterized as a long (L), short move (S), or zero (0) which occurs in the case of a swap. The zone location of the destination of each robot move is identified in the “zone” column. If the process chamber could be in more than one zone (e.g. a develop bowl in zone1X or2X, then1X/2X is entered in the master move table. It should be noted that PR stands for photoresist process chambers, PTi stands for pass through input chambers (e.g., pass throughposition9A,9B or9C), PTp stands for a pass through chambers positioned near the rear robot assembly40 (e.g. one or more support chambers165), BARC is a bottom anti-reflective coating process chambers, PR B/C stands for photoresist process bake and chill chambers, BARC B/C are bottom anti-reflective coating bake and chill process chambers, PEB/C are PEB bake and chill chambers, and DEV are photoresist develop chambers. In this example the Exposure process step (e.g.,step514 inFIG. 1F) is replaced with a drop at the pass through (Drop PTp) for therear robot assembly40.
• InFIG. 3A, the phasing of thesecond robot assembly11B is set up in a manner to enhance the limited throughput of the system. For this example, all of the single blade moves are scheduled to occur during the same “beat” withinmove #1 for each robot. This creates a total robot cycle time of 19.0 seconds which is equal to 189 WPH. For a 45 second process time, the effective throughput of the first chamber is equal to (3*3600)/64.8 or 167 WPH.
• FIG. 3B is another master move table1310 for a three robot system representing BARC process flow with maximum robot limited throughput. The phasing of thesecond robot assembly11B is changed to reduce the swap time at the PR bowl to 6 seconds. In this case, the throughput limit for the PR bowls is equal to (3*3600)/51 or 212 WPH. Therefore, changing the phasing does have an impact on the robot limited throughput. The total robot cycle time is 21.6 seconds which is equal to 167 WPH.
• Both schedules,FIG. 3A andFIG. 3B, produce the same system throughput of 167 WPH. However, if the bowl process is reduced from 45 seconds to 40 seconds then the bowl capacity is increased to 180 WPH and it is preferable to operate in the mode that provides the maximum robot limited throughput, which is the schedule inFIG. 3A. In this case, the throughput would again be limited by the chamber at (3*3600)/59.8 or 180 WPH which is preferable to the robot limited throughput case of 167 WPH inFIG. 3B.
• If the process time is increased to 50 seconds, then operating in the long swap condition would reduce the throughput below 167 WPH resulting in a throughput of (3*3600)/69.8 or 154 WPH. In this case it is best to operate in the mode that provides the maximum chamber throughput (including swap time). In this case, the throughput would be limited to 167 WPH.
• FIGS. 3A and 3B demonstrate the importance of establishing what the bottleneck resource is so the bottleneck can be properly managed. Proper management of the bottleneck resources helps avoid situations where small changes in recipe times lead to confusing throughput results.
• Offline Application
• Factories often use different methods and/or software to plan and schedule batches of substrates through the factory. These are often referred to as schedulers or dispatchers. These solutions generally implement a dispatch logic in which predetermined rules are used to define what action should be performed next by the equipment or the factory.
• The offline application provides optimal time-based chamber process recipes and/or process sequences without tying up the cluster tool. The offline application generally allows the user to model, analyze, and control the process variables within a process sequence in an unobtrusive manner to allow the throughput and/or process results received in one or more batches of wafers to be optimized. In one embodiment, a time-based sequencer is put on an off-line server to develop optimum process recipes and process sequences. The offline application can be used for recipe development and can display both a standard recipe and one or more modified recipes for comparison. The offline application can perform the following: download schedules from the cluster tool, up-load recipes and sequences to the cluster tool, and execute schedules. The offline application allows for modeling tool output (throughput of the tool) given the optimum processing requirements set by the user. The offline application allows automatic tact-time (wait time) calculations at each process step to allow for optimization of the processing recipe and process sequence recipe. The offline application allows the user to minimize the effect of planned delays and automatically resolves robot conflict to provide the highest robot-limited throughput. The offline application allows the user to schedule planned delays, such as maintenance activities. The offline application allows the user to decide when to start a trailing lot in a cascading mode (multiple lots run in series). The offline application allows the user to define desired process and system constraints when running a desired chamber process recipe or process sequence. The offline application allows the user to select schedules from proposed/modeled candidate schedules created by the controller. The offline application facilitates the selection of the optimal track system to use in the fab. The offline application modeling allows the user to decide if cascaded lots can be inter-woven. The offline application allows the user to schedule maintenance (e.g. bowl wash, calibration) based on when the module is scheduled to be idle. The offline application also allows the user to model “What if” scenarios (an exception occurs), and the possible/optimal recovery moves.
• Another feature of the offline sequencer is that the logic and analysis components required to set up and run a process sequence on multiple cluster tools can be made offline in a single system that is in communication with multiple “dumb” cluster tool controllers (e.g., no adaptive sequence type control), which just execute the sequence commands sent by the offline sequencer. Since the offline controller contains most of the logic components of the system this configuration can greatly decrease the cost of ownership and complexity of each of the multiple cluster tools.
• FIG. 4 is a process flow diagram1400 describing a method of using an offline application to create a schedule for a track system. Atstep1410, the method begins by pre-planning a schedule and the corresponding actions performed by each component of thecluster tool10. Instep1420, the schedule is stored on astorage device604. Instep1430, the schedule is transferred from thestorage device604 to acontroller101 in communication with thecluster tool10 where the schedule is stored.
• FIG. 5 is a process flow diagram1500 describing the pre-planning method ofstep1410 used to create a schedule for a track system. In one embodiment, the method may take the form of a question and answer interface that guides the user through the development of an optimal processing recipe and process sequence recipe. Instep1510, constraint conditions are defined by the user or system. Typical constraint conditions include, for example, those noted previously such as restricting use of a particular robot blade for a particular step, optimization for swap time for a process or for the maximum throughput, and optimization for constant wafer history or throughput.
• Instep1520, rules incorporating the constraint conditions are developed. Typical rules may include anomaly condition rules such as maximizing system level throughput if one chamber is down or retaining wafer history if one chamber is down. For example, if a system has a chamber type, referred to as “chamber type A,” with 3 chambers and a process time including wafer swap time of 90 seconds, the chamber type A bottleneck throughput is 120 wafers per hour. If the same system has a chamber type, referred to as “chamber type B,” with 3 chambers and a process time including wafer swap time of 70 seconds the chamber type B bottleneck throughput is 154 wafers per hour. If one of the type B chambers fails, the system could continue to process wafers at a rate of 103 wafers per hour (⅔* 154). However, the wafer history will change. Since the system initially has the same number of type A and type B chambers, the system could be re-configured to have one of the type A chambers disabled from operation so the system would run at a rate of 80 wafers per hour (⅔*120) with a wafer history similar to the original setup. If the system did not have the same number of type A and type B chambers, then removing one type A chamber would not rebalance the system. The system could also be configured to maintain all three type A chambers as operational with the two type B chambers, using the original beat frequency and schedule but invoking a ghost move on every third wafer removed from the cassette or wafer carrier. This ghost move would move a “phantom” wafer though the system, a wafer that is scheduled to be delivered at some point to the failed chamber B. This configuration would yield a throughput of 80 wafers per hour with the original wafer history. Normally, different rules are configured on tools as part of the system setup and apply to a recipe or are defined by the recipe. The offline application allows for all rules to be defined as part of the product recipe and modeled so the tool behaves exactly as intended by the engineering. Without this, specific tool behavior, especially during anomaly conditions, may not be optimized. The rules are used to control the various components in the system to meet the defined constraint conditions.
• Instep1530, a set of proposed schedules is developed. Instep1540, specific characteristics of each schedule are defined by analysis performed by the system controller. Instep1550, the system controller uses computer modeling to review and ensure that the proposed schedules are executable by thecluster tool10. The computer model contains real data, software, and all other aspects that will allow the evaluation of tool performance based on the defined constraints and the applicable rules. The computer model provides a process of developing the rules to meet a desired goal by simulating how the tool will perform based on different scenarios. The controller, through use of a software algorithm stored in memory is able to calculate the timing and movement of substrates through the system based on the rules, constraints, and actual input data. The specific characteristics allow the user to weigh the various benefits of selecting or prioritizing rules and/or constraint conditions in a desired order. Instep1560, one of the proposed schedules may be selected by the user and then uploaded to the cluster tool.
EXAMPLE
• Examples of the above introduced offline application are further described with reference toFIGS. 6-9.FIG. 6 illustrates an exemplary process flow diagram1600 describing an example of the rules referred to instep1520, used to, through software stored in the memory of a controller and some user input, create a schedule for a track system. Although the current example refers tocluster tool10 with a three robot configuration, it should be noted that the three robot configuration is used only for illustrative purposes, and the offline application and rules can be used to develop schedules for other chambers, for example, chambers containing one, two, or four robot configurations.
• Instep1605, the process begins by examining and inputting the process chamber layout, robot access zones and other system constraints into the offline application software. The input data may be in the form of 3-dimensional positional data of process chamber information (e.g., entrance information), robot movement, and motion control information (e.g., velocity position limitations, etc.). Next, instep1610 the required process flow steps for the schedule are entered into the offline application software. Next, instep1612, a decision is made as to whether the cluster tool configuration has the necessary hardware to meet the requirements defined by the constraints and the rules. If the necessary hardware is not present, the user will be alerted so that another process sequence can be selected so that the new tool configuration can be optimally used. If a new tool configuration is applied, instep1615, an estimate of the number of chambers required to perform the desired process sequence for the new tool configuration is made. If a new tool configuration is not applied,step1615 is skipped. Next, instep1620, the process steps are partitioned to the various cluster tool robots and the candidate transition points are defined. Next, instep1625, zones (regions) within the cluster tool and the moves which each robot will optimally make when the process sequence is performed are mapped for each robot using a Zone/Move table. Next, instep1630, tables containing all of the moves that each robot will perform are collected and placed in a master move table. Next, instep1635 the exchange time is recalculated to verify chamber capacity. If additional chamber capacity is required, additional chambers are added to the system instep1637 andsteps1620,1625,1630, and1635 may then be repeated. Next, instep1640, TACT time is distributed and bottle neck resources are defined.
• Instep1605, the module layout, zones, and constraints are examined. The following constraints are place on the architecture ofcluster tool10 in this illustrative example.Cluster tool10 is architected to provide a solution that meets product and processing requirements while also providing a solution that enables robust scheduling. In one example,cluster tool10 is designed with afirst processing rack60 and asecond processing rack80. Thefirst processing rack60 and thesecond processing rack80 contain one or more groups of vertically stacked processing chambers. These processing chambers are defined as follows: a coat bowl tower, a develop bowl tower, a thermal stack tower, afront end module24, an external module5 (e.g. stepper or scanner), and acentral module25. The central module contains a robot that moves wafers to a stepper interface robot and loads PEB chambers. Theexternal module5 contains arobot40 that moves wafers from the OEBR or stepper and back to the buffer. Thefront end module24 contains a frontend robot assembly15 that moves wafers from the output pass through back to the pod assemblies105.
• For this illustrative example, thecluster tool10 is designed as follows. The process chambers are designed with a size that ensures that the robots have adequate clearance from adjacent modules to reduce or eliminate collision avoidance. Only coat bowls will be located in thefirst processing rack60 of thecluster tool10. Only developer bowls will be located in thesecond processing rack80 of thecluster tool10. Coat bowl and developer bowls are not mixed. For this illustrative example, enforcing this constraint ensures that wafers start the process flow on the left side of the cluster tool10 (adjacent first processing rack60) and once the wafer leaves the left side of thecluster tool10, the wafer does not need to return to that side. This architectural decision creates a “quasi-parallel” flow that simplifies the scheduling and architecture with little or no negative impact. For the three robot system, the robots are named as follows: thefirst robot assembly11A (left), thesecond robot assembly11B (middle), and thethird robot assembly11C (right). Pass throughpositions9A-9B, accessible by thefirst robot assembly11A and thesecond robot11B are installed at the input of thecluster tool10. Pass throughpositions9B-9C, accessed bysecond robot11B andthird robot11C, are installed at the output of thecluster tool10.
• As noted above,FIG. 7 is a schematic plan view of asimplified version1700 of thecluster tool10 ofFIG. 1 showingdifferent chamber zones1705,1710,1715,1720,1725, and1730. These chamber zones are defined to simplify scheduling and eliminate collisions. Chamber zone1 (1705) represents the first coat bowl stack and also includes pass throughposition9A infirst processing rack60. Chamber zone2 (1710) represents the second coat bowl stack in thefirst processing rack60. Chamber zone3 (1715) represents the thermal stacks in thefirst processing rack60. In another embodiment multiple zones are used to define the thermal stacks in thefirst processing rack60. Chamber zone1X (1720) represents the first develop bowl stack in thesecond processing rack80 and includes pass throughposition9C. Chamber zone2X (1725) represents the second develop bowl stack in thesecond processing rack80. Chamber zone3X (1730) represents the thermal stacks in thesecond processing rack80. In another embodiment multiple zones are used to define the thermal stacks in thesecond processing rack80. In general, when one of thefirst robot assembly11A, thesecond robot assembly11B, or thethird robot assembly11C is accessing a chamber zone, the other two robots are locked out from moving in the same zone. In general, all three robots move to different chamber positions within different chamber zones at the same time and will also initiate transfer moves (pick/drop) at the same time, this is referred to as the “beat frequency,” which is described above. In some cases, wafer throughput can be increased by allowing thefirst robot assembly11A to complete a single blade move/pick or move/drop whilesecond robot assembly11B andthird robot assembly11C are scheduled to complete a swap move on the right side of thecluster tool10. Similarly, thethird robot assembly11C could freely move in a chamber zone if thefirst robot assembly11A and thesecond robot assembly11B are accessing the right side of thecluster tool10.
• Although not addressed in this illustrative embodiment, in another embodiment, it is possible for two robots of thecluster tool10 to access the same zone at different heights, however, because of potential collisions at some locations, certain move sequences are prevented.
• Instep1610, the required process flow steps for the schedule are defined. The process flow steps are determined by the process engineer and the needs of the customer. First, the basic flow process must be converted into a process flow that matches the specified cluster tool architecture. This involves the following steps for this illustrative example which could be modified as the cluster tool architecture changes. The purpose of thisstep1610 is to extract only the steps that impact the process unit and the scheduling of the robots. Other steps, namely those covered by the frontend robot assembly15, therear robot assembly40, and the stepper interface robots (not shown) are not included in this exemplary schedule. The exemplary baseline process flow used for this analysis involves Basic Organic BARC. The initial sequence recipe includes the following steps: BARC—BARC B/C—PR—PAB/C—Exposure—PEB/C—Dev.
• The Basic Organic BARC process flow is modified for thespecific cluster tool10. First, pass through steps are added at thefront end module24 for pass through input (PTi) and for the pass through output (PTo). The Exposure step is replaced with a drop at the pass through (Drop PTp) for therear robot assembly40. The PEB/C step is replaced with a Pick from the PEB module defined as Pick PEB/C which is the standard flow for PEB. The modified process flow for Basic Organic BARC is as follows: PTi—BARC—BARC B/C—PR—PAB/C—Drop PTp—Pick PEB/C—Dev—PTo.
• Instep1615, an estimate of the number of chambers required for the modified process is made. This is a first approximation using the defined process times and the estimated chamber exchange times. The number of chambers required is calculated as follows:
• Chambers=(Stepper Throughput (Chamber Process Time+Exchange Time))/3600
• Stepper Throughput=Wafers/Hour (or desired throughput of the system)
• Chamber Process Time=Process time in seconds
• Exchange Time=exchange time in seconds
• Chamber exchange times may be determined as follows. For chambers designed for swapping (Coat, Develop, and Swapping B/C/C), the robot swap time is used. A “swap” equals pick time plus drop time. For chambers designed for single blade Pick/Drop (Pass through chambers, Single Blade B/C/C) the single blade time is used. Thus, for single blade Pick/Drop chambers, the exchange time equals pick time plus move time plus drop time plus move time plus pick time plus move time plus drop time. A Long Move (ML) is defined as the longest move time (bottom left to top right within the tool), in this case the ML takes approximately 1.8 seconds. In one example, the time it takes a robot to move between chambers may vary from about 1.0 seconds to about 1.8 seconds. A swap process where a robot picks up and drops a substrate, “quasi-swap”, may be about 2.1 seconds each. A short move (MS) is defined as a short move in the same process module stack, in this case for a “quasi-swap” of the single blade B/C/C module takes approximately 1.0 seconds. A “quasi swap” involves a pick/drop followed by a short move followed by a drop/pick. A Pick (P) takes about 2.1 seconds and a Drop (D) takes about 2.1 seconds.
• The concept for the current chamber architecture ofcluster tool10 assumes that thefirst robot assembly11A, thesecond robot assembly11B and thethird robot assembly11C are all in motion at the same time between chamber positions and Pick/Drop transfer steps are completed at the same time without movement of the major robot components (e.g.horizontal assembly90, vertical assembly95). As a result, the worst case move time is used for defining the schedule. In one embodiment, the opportunity exists to shorten the joint move time if all robots are scheduled for a shorter move. However, for this illustrative example, only the worst case move is considered for the long move and the 1.0 second move time for the short move within the same stack.
• Instep1620, the process steps are partitioned to a robot and the candidate transition points are defined. The current robot architecture ofcluster tool10 is designed to share the load during the sequential processing of the substrates as equally as possible among the process robots. At least two of the process robots can reach each chamber position and as a result, the processing steps assigned to each robot can be modified in order to balance the load. In all cases, the process flow is distributed to the robots with thefirst robot assembly11A processing the initial steps, thesecond robot assembly11B processing the middle steps, and thethird robot assembly11C processing the final steps of the process. This assignment ensures that once a robot has completed its work with a given wafer, the wafer never unnecessarily returns to that robot, which can waste time and hurt throughput. This eliminates the complexities associated with re-entrant flows of wafers to the same robot. The steps used to partition the flow are defined below.
• Balancing involves optimizing throughput by distributing the load among the robots. Balancing of the robot load is divided into three categories. First, for “balanced flows,” for example, 3, 6, 9, 12 . . . steps, the moves are distributed evenly to each of the three robots. For “unbalanced heavy flows,” for example, 5, 8, 11, 14 . . . steps, the outside robots,first robot assembly11A andthird robot assembly11C perform the same number of moves and thesecond robot assembly11B is loaded with one less move. Since thesecond robot assembly11B can access both sides, it is best to only maximize the loading on this “flexible” resource when absolutely necessary. For “unbalanced light flows, for example, 4, 7, 10, . . . steps, moves for thesecond robot assembly11B and thethird robot assembly11C are balanced and thefirst robot assembly11A performs one additional move.
• After the robot loads are balanced, the transition points for each robot are defined using the following criteria. The input and output transition moves for each robot sequence is a single blade move. The input is a single blade Move/Pick the output or last step is a single blade Move/Drop. While partitioning the process moves, these can be considered as half moves. For a single blade B/C/C performing this exemplary recipe the robots act as follows: thefirst robot assembly11A: Pick PTi—swap BARC—Drop/Pick BARC B/C—Drop PR. For thesecond robot assembly11B the steps are as follows: Pick PR—Pick/Drop PAB/C—Drop PTp. For thethird robot assembly11C the steps are as follows: Pick PEB/C—swap Dev—Drop PTo. For a swappable B/C/C performing the exemplary recipe the robots act as follows: for thefirst robot assembly11A the steps are as follows: Pick PTi—swap BARC—swap BARC B/C—Drop PR. For thesecond robot11B the steps are as follows: Pick PR—swap PAB/C—Drop PTp. For thethird robot assembly11C the steps are as follows: Pick PEB/C—swap Dev—Drop PTo.
• Instep1625, zones and moves are mapped for each robot using a Zone/Move table.FIGS. 8A-8C are robot move tables for thefirst robot assembly11A, thesecond robot assembly11B, and thethird robot assembly11C respectively. With reference toFIG. 8A, the move table1810 for thefirst robot assembly11A is created as follows. First, the start moves are completed. The first step is always a PTi for both the Move (M) and the Pick (P). The chamber position Move (M) is identified as Long (L) or Short (S). The robot zone location for each move is also identified. If the chamber could be in more than one zone (e.g. a develop bowl in zone1X or2X, then1X/2X is entered. Next, the first step is completed. The current example is for a single blade BARC flow with single blade B/C/C. For the chamber position moves (M) identify if the move is long (L), short (S) or zero (0) in the case of a swap. The sequence for a swap is Move/Pick/Drop and the line for the second chamber move (M) in the table is left blank and no step name would be placed in the position. Next, the middle moves are completed. In this example, the middle moves include a move to the single blade B/C/C. After the middle moves are complete, the last half move is the drop at the transition chamber, which in this case is the PR (photoresist or coater) chamber. The same process was completed to develop the move table1820 for thesecond robot assembly11B,FIG. 8B, and the move table1830 for thethird robot assembly11C,FIG. 8C
• Instep1630, a master move table is built.FIG. 9 illustrates a master move table for the three robot system of thecluster tool10. The master move table is similar to the robot table, however, the start and end moves for thefirst robot assembly11A (Move PTi/Pick PTi and Move PR/Pick PR) are “wrapped” together to combine these two half moves into one single substrate move sequence at the top of the master move table. In this step of the process, the individual robot moves are placed into a common schedule table. Rules are applied to adjust the phasing of the robots to avoid zone conflicts and to group moves to optimize the flow. “Phasing” involves aligning the moves of each robot. For example, all three robots perform a pick/drop or drop/pick during the same “beat” or all three robots move to different chamber zones during the same “beat.” Grouping the moves and adjusting the phasing provides opportunities to reduce the time for a complete robot cycle time from the maximum time (all single blade moves) by completing “quasi swap” or “swap” moves at the same time across all the robots.
• Instep1635 the exchange time is recalculated to verify chamber capacity. The exchange times for each chamber are recalculated to determine whether the exchange time is different from the estimated exchange times fromstep1615. This can be done directly from the master move table ofFIG. 9, assuming that the system is operating at robot limited throughput. If the exchange times are different, then the number of chambers required is updated instep1637. A difference in exchange times could be caused if the phasing adjustment to avoid collisions causes a shift at a transition module resulting in a longer exchange time for that module. The data in the master move table can also be used to calculate the bottle neck resources. If the robots are free to operate at a lower speed because of some other bottleneck, the bottleneck condition is defined.
• Instep1640, TACT time is distributed and bottleneck resources are defined. If fortunate enough to have extra resources (i.e. chamber or robots not in the bottleneck), then the extra time (spare utilization) may be distributed as TACT time. Various rules for optimization of extra time will apply based on processing requirements and specific customer preferences.
• TACT time is directly related to the wafer capacity. For example, if thecluster tool10 is running at 180 WPH or 20 second cycle time, the process time plus the exchange time plus TACT time will add up to some multiple of 20 seconds. The exchange time is determined by how the flow (wafer in to wafer out) is run. The distribution of TACT time depends on the processing requirements and specific customer preferences. For example, referring to the master move table ofFIG. 9, if the BARC process time is 45 seconds and the exchange time is 4.2 seconds, then TACT time is 13.2 seconds (e.g. 20.8*3=62.4; 62.4−45−4.2=13.2 seconds).
• Cascadinq
• As noted above one of the key elements to a desirable process sequence performed on a substrate in a cluster tool system is the need to assure that each wafer processed has a similar wafer history while also maintaining a high wafer throughput. This key element is also applicable to the processing of wafer lots. On most track lithography systems, several substrate processing constraints must be satisfied to assure that a constant wafer history is achieved. These constraints are usually satisfied by creating large gaps between the lots of wafers. Unfortunately, these large gaps lead to an increase in lot processing time and a corresponding decrease in the efficiency of the track system. While it can be argued that changes in wafer history at certain points in the process flow will have little or no impact on the process flow, whether these changes impact the process flow is influenced by several factors including: the type of resist, the device, and the process window for each specific process. However, it is still preferable to maintain a constant wafer history.
• Another process for increasing the system throughput while maintaining constant “wafer history” and reducing the system cost and complexity is called “cascading.” Cascading is defined as running two or more lots in a track system at the same time, for example, lot A and lot B. The goal of cascading is to provide the user with the capability to minimize the gaps between lot A and lot B and thus increase substrate throughput, while maintaining a constant wafer history for the wafers in each lot. Minimizing the gaps between lot A and lot B increases the utilization of the scanner and thus the productivity of the cluster tool.
• In certain embodiments, in order to maintain the same wafer history, several criteria must be met. First, the cycle time from the pass through input (PTi) to thescanner interface5 must be the same for all wafers within a lot, except for wafers that have completed PAB and are waiting for thescanner5. Due toscanner5 processing variability, it is sometimes necessary and acceptable to buffer wafers in front of thescanner5. Second, cycle time from exposure completion to the beginning of post exposure bake (PEB) start—Post exposure delay (PED)—is set as a track recipe parameter. Third, the cycle time from PED chill to pass through output is repeatable within 3 seconds. Wafers exit the scanner at regular intervals and as a result, these wafers will enter the PEB at regular intervals. However, the wafers must leave the PEB module at regular intervals consistent with the Process Transfer Robot Schedule. The only way to match the irregular flow with the regular flow is to allow variable time (buffer time) at the PEB chill step.
• Wafer history must be maintained in order to achieve repeatable process results. This means that the average throughput, robot timing, and schedule rules, must be applied in the same manner. For example, if lot A is running on the tool with one recipe and lot B is scheduled next, but has a shorter recipe time for most steps, adding additional slack time to the lot B process time is not permitted as a means to improve cascading.
• There are several possible scheduling modes or embodiments for cascading. The first scheduling mode is a more conventional schedule in which the robots and slack time are matched as much as possible with the scanner throughput and/or chamber bottleneck(s). In the case of the first scheduling mode, slack time for the chambers and robots is evenly distributed so the system runs at a slower rate. The second mode is a constant robot throughput schedule that maximizes the opportunity for cascading. In the case of the second scheduling mode, the system runs at a faster rate and the slack time is lumped as a wafer send skip or “phantom wafer.”
• The first scheduling mode creates an enhanced schedule for a given product based on the scanner throughput and process recipe independent of other lots. The robots and other system resources are balanced with the scanner throughput. The robot synchronization timing is set to match thescanner5. In most cases, the process robots will have slack time. Thus the robots will be scheduled to have increased translation or move times from module to module. In one embodiment robot move time is increased by forcing the robot to wait before doing a drop. In another embodiment, robot move time is increased by reducing the speed of the robot.
• In the first scheduling mode, the scheduled throughput of the track is slightly faster than thescanner5. If the track gets ahead of thescanner5, wafers will begin to fill the scanner interface buffer. Only then will the track skip a wafer send or several wafer sends, to force reduction in the buffer.
• With the first scheduling mode, in certain embodiments, the same schedule may be run each time the same recipe sequence and product is run on the tool. If a first scheduling mode schedule is run on the tool for lot A, then lot B will have to wait until all wafers from lot A have reached the PEB before the first wafer from lot B is picked from the input pass through. This creates a gap in the tool sequence pipeline that ensures that no resources are serving both lot A and lot B simultaneously. In some embodiments, lots are cascaded with a small gap in the pipeline.
• The second scheduling mode creates an enhanced schedule for a given product based on process recipe and scanner throughput in much the same manner as the first scheduling mode, however, in the case of the second scheduling mode, the schedule is defined to maximize cascading flexibility. This is done by running all process robots at the same maximum robot limited throughput for all process flows. In the first scheduling mode, slack time for the robots and the process modules is distributed throughout the schedule to closely match the running rate of the track with the scanner. The track will skip a wafer send only when the buffer begins to fill at the scanner interface. In the second scheduling mode, the track speed is not matched with the scanner speed but rather is higher than the scanner speed, possibly much higher. As a result, the scheduler proactively implements skips in sending wafers so the average track throughput is equal to or slightly higher than the scanner throughput. Additional wafer send skips are implemented if the buffer begins to fill beyond the send limit.
• In the second scheduling mode, all wafers will see the same schedule with minimal slack time. Slack time is eliminated or minimized because each wafer is sent through the tool at maximum speed (minimum cycle time). Average throughput to the scanner is controlled by skipping a wafer send. For example, if the maximum speed of the track is 10% higher than the stepper, than after every 10 wafers, a wafer send will be skipped. This skipped wafer is scheduled through the system as a “phantom wafer.” As a result, the robots continue to move as if they were physically moving the phantom wafer. In some embodiments, it is necessary to randomize or shift the skip point to rotate through the process modules. For example, if three bowls are active and the track needs to skip every third wafer send to match the scanner speed, then the same bowl will always be skipped. In this case a decision must be made to use only two of the bowls for the process or to shift the wafer send skip so the idle time is rotated to each bowl.
• FIG. 10A is atime chart2000 representing the time it takes to process three lots of 25 wafers each using techniques known in the prior art. The x-axis represent the processing time in seconds. The y-axis represents the wafer number in each of the three lots, for example, wafers 1-25 for each of a first, second, and third lot respectively. This example was performed by running three different lots of 25 wafers each. The goal is to complete all three lots within the fastest time while maintaining a constant wafer history. In this prior art example, three lots are run without cascading. The flows of the wafers for each lot are mapped out showing no resource conflict and thus a steady-state wafer flow. The processing of the last wafer,wafer #25 oflot1, is completed beforewafer #1 oflot2 begins processing. The processing of the last wafer,wafer #25 oflot2, is completed before the processing ofwafer #1 oflot3 begins. The total amount of time to run all three lots is the sum of each of the lots running separately. The total time for completion is 2310 seconds.
• FIG. 10B is atime chart2010 representing the time it takes to process three lots of 25 wafers each using one embodiment of the present invention. The x-axis represents the processing time in seconds. The y-axis represents the wafer number in each of the three lots, for example, wafers 1-25 for a first, second, and third lot respectively. Using the cascading method,wafer #1 oflot2 begins processing while some wafers fromlot #1 are still in the tool.Wafer #1 oflot2 begins processing approximately 18 seconds afterwafer #25 oflot1 begins processing.Wafer # 1 oflot3 begins processing while some wafers fromlot2 are still in the tool.Wafer #1 oflot3 begins processing approximately 16 seconds afterwafer #25 oflot2 begins processing. The graph shows no resource conflict during the transition between each flow. The use of the current method allows for the completion of 3 lots of wafers in about 1790 seconds. That is a savings of 520 seconds while maintaining a constant wafer history for each lot.
• While the foregoing is directed to embodiments of the present invention, other and further embodiments of the invention may be devised without departing from the basic scope thereof, and the scope thereof is determined by the claims that follow.

Claims (22)

1. A method of improving substrate throughput prior to processing a plurality of substrates in a cluster tool having a plurality of processing chambers and at least two robots, comprising:

planning a schedule and the corresponding actions performed by each component of the cluster tool;

storing the schedule on a storage device; and

transferring the schedule from the storage device to a controller in communication with the cluster tool.

2. The method ofclaim 1, wherein the planning a schedule comprises:

defining constraint conditions;

developing rules incorporating the constraint conditions;

developing a set of proposed schedules;

defining specific characteristics of the proposed schedules;

ensuring that the proposed schedules are executable by the cluster tool; and

selecting at least one of the proposed schedules.

3. The method ofclaim 2, wherein the developing rules incorporating the constraint conditions comprises:

defining process flow steps for scheduling and required throughput;

estimating a number of processing chambers required within the cluster tool;

partitioning the process flow steps to the at least two robots;

defining candidate transition points within the cluster tool;

mapping out chamber zones and moves for the at least two robots;

building a master move table;

recalculating the exchange time to verify chamber capacity;

distributing slack time; and

defining bottle neck resources.

4. The method ofclaim 1, further comprising controlling the transfer of a plurality of substrates through the plurality of processing chambers according to the planned schedule.

5. The method ofclaim 1, further comprising depositing one or more uniform resist layers on the surface of the first substrate.

6. The method ofclaim 5, further comprising transferring the first substrate out of the cluster tool to a separate stepper or scanner tool to pattern the substrate surface by exposing the resist layer to a resist modifying electromagnetic radiation and then developing the patterned resist layer.

7. The method ofclaim 4, wherein the controlling the transfer of a plurality of substrates through the plurality of processing chambers according to the planned schedule comprises:

transferring a first substrate within the cluster tool using a first robot assembly, wherein transferring the first substrate includes removing the first substrate from a first position and placing the first substrate in a second position;

transferring a second substrate within the cluster tool using a second robot assembly, wherein transferring the second substrate includes removing the second substrate from a third position and placing the second substrate in a fourth position;

synchronizing the removing the first substrate with the removing the second substrate or placing the second substrate during a first period of time; and

synchronizing the placing the first substrate with the removing the second substrate or the placing the second substrate during a second period of time, wherein the first period of time and the second period of time are distinct and non-overlapping.

8. The method ofclaim 2, wherein the selecting at least one of the proposed schedules further comprises displaying the proposed schedules for comparison with a standard schedule.

9. The method ofclaim 4, wherein the controlling the transfer of a plurality of substrates through the plurality of processing chambers according to the planned schedule comprises:

introducing a first lot of wafers containing one through n-th wafers into a cluster tool containing one or more processing chambers;

processing the first lot of wafers for a first time period;

introducing a second lot of wafers containing one through n-th wafers into the cluster tool prior to completion of the first time period, wherein the second lot is introduced so as to minimize a time gap between the n-th wafer of the first lot of wafers and the first wafer of the second lot of wafers while maintaining a first constant wafer history for each wafer within the first lot and maintaining a second constant wafer history for each wafer in the second lot; and

processing the second lot of wafers for a second time period during at least a portion of the first time period.

10. The method ofclaim 1, further comprising:

measuring metrology data for processing a lot of substrates on the cluster tool prior to planning a schedule.

11. The method ofclaim 1, wherein planning a schedule further comprises resolving robot conflict to provide the highest robot limited throughput.

12. A method for improving substrate throughput prior to processing a plurality of substrates in a cluster tool having a plurality of processing chambers and at least two robots, comprising:

developing a process sequence, wherein the process sequence comprises:

depositing one or more uniform resist layers on the surface of the plurality of substrates;

transferring the plurality of substrates out of the cluster tool to a separate stepper or scanner tool to pattern the surface of the plurality of substrates by exposing the resist layer to a resist modifying electromagnetic radiation, and then

developing the patterned resist layer

storing the process sequence on a storage device; and

transferring the process sequence from the storage device to a controller in communication with the cluster tool.

13. The method ofclaim 12, wherein the developing a process sequence comprises:

defining constraint conditions;

developing rules incorporating the constraint conditions;

developing a set of proposed schedules;

defining specific characteristics of the proposed schedules;

ensuring that the proposed schedules are executable by the cluster tool; and

selecting at least one of the proposed schedules.

14. The method ofclaim 13, wherein the developing rules incorporating the constraint conditions comprises:

defining process flow steps for scheduling and required throughput;

estimating the number of processing chambers required within the cluster tool;

partitioning the process flow steps to the at least two robots;

defining candidate transition points within the cluster tool;

mapping out chamber zones and moves for the at least two robots;

building a master move table;

recalculating the exchange time to verify chamber capacity;

distributing slack time; and

defining bottle neck resources.

15. The method ofclaim 13, wherein defining constraint conditions further comprises minimizing the effects of planned delays.

16. A system for processing a substrate, comprising:

a cluster tool having a plurality of processing chambers and at least two robots;

a controller in communication with the cluster tool;

a memory coupled to the controller;

an offline server;

a storage device coupled to the offline server; and

a memory coupled to the offline server, the memory comprising a computer-readable medium having a computer-readable program embodied therein for improving the throughput of substrates in the cluster tool, the computer-readable program comprising:

computer instructions for developing a recipe for improving substrate throughput in the cluster tool by controlling the movements of the at least two robots, the computer instructions comprising:

i. planning a schedule and the corresponding actions performed by each component of the cluster tool;

ii. storing the schedule on the storage device; and

iii. transferring the schedule from the storage device to the controller.

17. The system ofclaim 16, wherein the computer instructions further comprise:

defining constraint conditions;

developing rules incorporating the constraint conditions;

developing a set of proposed schedules;

defining specific characteristics of the proposed schedules;

ensuring that the proposed schedules are executable by the cluster tool; and

selecting at least one of the proposed schedules.

18. The system ofclaim 16, wherein the schedule is transferred from the storage device to the controller in communication with the cluster tool.

19. The method ofclaim 17, wherein the developing rules incorporating the constraint conditions comprises:

defining process flow steps for scheduling and required throughput;

estimating the number of processing chambers required within the cluster tool;

partitioning the process flow steps to the at least two robots;

defining candidate transition points within the cluster tool;

mapping out chamber zones and moves for the at least two robots;

building a master move table;

recalculating the exchange time to verify chamber capacity;

distributing slack time; and

defining bottle neck resources.

20. The system ofclaim 16, wherein the cluster tool comprises:

a first processing rack containing a first vertical stack of substrate processing chambers;

a first robot adapted to transfer a substrate to a substrate processing chamber in the first processing rack;

a second processing rack containing a first vertical stack of substrate processing chambers; and

a second robot adapted to transfer a substrate between a substrate processing chamber in the first processing rack and a substrate processing chamber in the second processing rack.

21. The system ofclaim 20, wherein the substrate processing chambers are each selected from one of the following: a coater chamber, a developer chamber, a HMDS chamber, a chill chamber, or a bake chamber.

22. The system ofclaim 16, wherein the cluster tool further comprises:

a third robot adapted to transfer a substrate among the substrate processing chambers within a first time period.

Images